Two months ago, I jumped into forex with zero experience—just ChatGPT and a lot of hope. After countless failures (slippage, overfitting, emotional trading), I finally cracked a method that’s working: Instead of hunting for “perfect” setups, I just exclude the worst ones.
My Breakthrough: Avoiding Losses, Not Predicting Wins
Most strategies fail because they overfit—they look amazing in backtests but explode live. My fix? Use ChatGPT to filter out bad conditions (like high-impact news, low liquidity, or choppy price action).
Risk-to-Reward: 1:4 – This means even a 50% win rate is profitable. My current stats: 198 wins, 178 losses (52% win rate).
Full Automation: I connected TradingView to MetaTrader, so the strategy runs itself (no emotional mistakes!).
How It Actually Works:
Step 1: Eliminate Bad Environments
Step 2: Strict 1:4 Risk-Reward
Step 3: Full Automation
What I want to do next is to make more strategies for different pairs to see if perfection is achievable.
Let me know if you tried before and what was outcome.
Drop me a text if you want to collaborate and share success
It’s unfinished. As I mentioned my target is 65-70%. Once it achieved I will run it with metatrader and be able to link and share. Only live trading I have at this stage is my laptop running this strategy 24/5 and see it performance live sort of. And it doesn’t disappoint for now
Forex. 30m timeframe. What I do from start is. Get some sort of strategy which act as base, just to generate as many trades as possible. Great if it has profit factor 1.0. Normally it’s between 0.8-1.0 and win rate comes around 20%. Once I’ve got it, instead of trying to figure out favourable conditions of those winning trades, I do opposite, asking Chatgpt to provide filters to exclude unfavourable conditions and keep everything else. In my case with this EURJPY strategy my starting point was somewhere about 3000 total trades and 520 winning ones. And as today after couple weeks out of it I’ve got 196 wins and 154 losses. Still some way to go to final target.
Anything above a 20% win rate is profitable, with an R of 4.
Truly amazing. You’re doing better than 99.99% of the world’s professional fund managers, and better than any other retail trader in history.
My thinking, also.
You do understand why everyone who replies is asking for some evidence, @Kylji ? Especially when you’re inviting people to “drop you a text” if they want to “share success”?!
Obviously sounds dodgy. My problem is time. The strategy I’m on EURJPY. And it’s only half. Long only. It is still unfinished, I already spent just on this part over two weeks, hopefully I’ll finish it by end of this week. And I’m working on it 10-12h a day. Once It’s completed on average it will be doing 2-3 trades per week. Once short only completed and they merged I expect to have about 1-2 trades a day, which is ok. So if let’s say you got 6 trades per week, you win 4, lose 2, which is not big return in short time. To make it good, I need 5 strategies like that. To make it on my own will take 6 months. If I go for win rate of 40% it is profitable and great, but, having 5 strategies with 40% win rate and running simultaneously will result in insane drawdowns.
I can share my latest project for GBPNZD, which was fine and still fine, but when I reached 65% win rate i noticed one issue and once I fixed it actual win rate dropped to 40%, however with 40% win rate it is still very good. As new user I cannot share link or upload file and code itself over 2000 lines and doesn’t fit in the message. Let me know if you got any workaround to share and I’ll be happy to
It sounds like it’s curve-fitted or not taking in to factor variable spread.
If I use MT5 Strategy Tester in OHLC/Calculated bars, I’ll get much better results than using Real Ticks from my broker’s Live feed.
I would say partially. It build for one pair on certain timeframe. My approach is take a pair, do in two parts - short and long as separate strategies. And after in case of GBPNZD you focus on volatility filters first, once you think you tried everything associated with volatility you move to something else. And to avoid overfitting you use AND NOT, which excluding anything unfavourable to the entry
